The Learning And Evolution Model Of Artificial Life And Petri Net Modeling

The Artificial Life, Alife for short, is one of newly interdisciplinary science fromthe late1980s. It mainly takes the comprehensive and bottom-up approach in thestudy, analyses the global evolution and the whole characteristic through theadjustment of interactive rules between the Alives.Learning and evolution is one of the most important characteristics of Alifedifferent from other Intelligence Science. The Alife system is a typical complexadaptive system (CAS). Every adaptive agent interacts with the environmentconstantly, and adapts itself to the environment according to the information observed.The mainly components of environment around agent are other adaptive agents; anybehavior taken will affects other agents nearby. While learning and evolution, Alivesselect the behavior based on their observation and experience, interact with otherAlives. The behavior would affect others’ observation, then affects others’ experienceand behavior consequently. As a result, a global dynamic will emerge bottom-up fromthese interactional affections, and lead to the realization of adaptive learning andevolution.Petri net is a graphical and mathematical modeling tool applicable to informationprocessing systems which are characterized as being concurrent, asynchronous,distributed, parallel or/and stochastic. As a graphical tool, Petri net has the advantageof visual-communication and easy using. Petri net not only can be used to depict thestructure, but can describe the dynamic activities of systems. Many researchersimproved it to many other field. This paper represents the Alife’s learning andevolution model by meanings of DSPN and GSPN, analysis the validation of themodels.The main work and innovation in this paper as follow:1. Based on the game learning theory, this paper adds the memory into Alife’slearning model, and proposes an experience weighted and limited memory learningmodel (ELM), studies the effects of learning when the memory would be forgotten. Inaddition, we simulated some kinds of game models on the swarm platform, observerthe result of learning.2. Based on the classified system and ELM, this paper adds anticipation and irrationalness into Alife’s learning model, and proposes an experience weighted andlimited memory evolution model (LMEM). In addition, we analysis the publicproblem model based on LMEM, observed the difference whether the Alifes couldanticipate based on swarm platform.3. By meanings of GSPN and DSPN, this paper represents the Alife’s learning andevolution model, memory, and active status, analysis the validation of the ELM andLMEM.4. This paper finally simulates a international oil market, instructs the action ofproducers and consumers by LMEM, shows the change of supply, demand and pricein the market. LMEM has been validated through the experimental results ofsimulation.